AI’s Impact on Law: Why the Transformation Narrative Is Overstated
A common message we hear today is that AI will soon bring about sweeping changes to the practice of law, making us so much more efficient that we’ll have plenty of time for other things.
I’ve kept my finger on the pulse of AI since ChatGPT appeared in 2022. I’m a heavy user of AI as a law professor and part-time criminal lawyer. I’m constantly experimenting with it and dazzled by its capabilities—you won’t find a bigger fan of AI.
But if there’s one thing that’s clear, it’s that AI will not transform the practice of law. Far from it.
Saying so paints a grossly misleading picture of the future of law—premised on a skewed reading of the facts and a misconception of how most lawyers spend most of their time.
The overheated claim
The claim we’re often hearing is crystallized in a recent post on Jordan Furlong’s (otherwise terrific) Substack titled “How will you spend your AI Dividend?”, where he writes: “But if there’s one effect of Gen AI that hardly anyone disputes, it’s that it vastly reduces the amount of time and human effort required to complete many standard legal tasks.”
True, but what does this mean for the practice as a whole? For Furlong, it means a “productivity revolution is underway in the law, and nobody knows how far it will take us.” His evidence:
“We’ve been aware of this since 2023, when a Legal Value Network study conservatively estimated that 20% of lawyer work tasks at large law firms could be performed by AI. Last week, Clio’s 2024 Legal Trends Report went further and concluded that 57% of lawyers’ hourly billable work activities could be automated by AI. This week, Thomson Reuters’s 2024 Future of Professionals Report predicted that AI could free up work time at a pace of four hours per week (200 hours annually) within one year, tripling that total within five. Want a real-life example? In the first quarter of 2024, Husch Blackwell displaced about 6,000 lawyer hours using AI tools.”
Drawing on these stats, Furlong asserts that “You could spend hundreds of thousands of dollars annually to support a bunch of leveraged lawyers grinding out billable work — or, you could replace most of them with a powerful Gen AI system…”
All of this is so wildly inaccurate — so patently implausible — I hardly know where to begin.
Misconstruing the evidence
The sources that Furlong cites point to efficiencies on two fronts. Using AI helps lawyers speed up document review and legal research.
Now just stand back and think about it: do lawyers at even the biggest firms spend so much of their time doing these things that 57% of their billable hours could disappear through automation?
Anyone who has worked in a law firm of any size could readily see why this is pure fiction.
But didn’t Husch Blackwell report that it displaced about 6,000 lawyer hours using AI tools? Isn’t that conclusive?
A closer look at the facts
Yes, Husch Blackwell did say that. But it also said that it has “more than 1,000 attorneys and nearly 1,200 staff members and paralegals.”
A thousand lawyers each billing, say, 1,500 to 2,000 hours a year amounts to 1.5 to 2 million billable hours a year. Six thousand hours is maybe the work of 3 or 4 lawyers, out of a thousand.
I concede that, at the extreme, AI might make a significant dent in the time a junior associate might spend wading through thousands of documents doing due diligence on a big corporate deal.
I remember doing this as an articling student at a large commercial firm on Bay Street. I also did a lot of research, and AI might have saved me some time there as well.
But this is not lawyering in the vast majority of cases, and even where it does fit this model, AI is not making 57% of the work disappear.
What practice really involves
I’ve had the privilege of working on files at a national firm for clients as big as Bell Canada and the Royal Bank, and later, at a small criminal firm, where at one point the client was a 12-year-old girl. I’ve been Crown and defence. I’ve assisted plaintiff and defendant in all kinds of cases.
I saw time and again that no matter how big or small the file, the general the nature of the work that lawyers do on a file is the same. And almost all of it requires a human-in-the-loop.
Whether you’re acting for Bell or a mom-and-pop shop, or a child, most of the time spent on a matter involves the same few tasks: holding the client’s hand, communicating with opposing counsel, and carefully reading and rereading crucial documents in the case.
Yes, you may do some research or document review — sometimes a lot. But in each of these cases, AI can only take you so far.
AI can save time combing through piles of documents or cases to find the ones that matter. But most lawyers spend more of their time pouring over or crafting key documents — contracts, pleadings, opinions — pondering strategy, implications, or advice.
And while AI might make finding law easier, you can only find things if you know what you’re looking for. Without a good grasp of law, you won’t know how to prompt AI to look in the right places. AI won’t replace our intuition. It can’t come up with creative ways to frame a legal claim.
AI will help us do many things. But the vast majority of our time on most files will still be spent on tasks we need to do ourselves: reassuring the client, persuading a judge or jury by reading the room, or reaching a settlement by using humility and common sense.
That’s the practice. It looked that way a hundred years ago. It will look mostly the same in a hundred more.
I would suggest that the “definition” of AI be outlined in an article in the beginning. Otherwise, we are talking about apples and oranges – two separate categories.
There’s a distinction that often gets lost when we contrast perspectives about the potential magnitude of impact on legal practice from AI. It seems to surface out of the contrasting starting points about using AI to do existing work more efficiently (through some kind of “easy button” adoption of a packaged service) versus leveraging AI to drive real practice changes. The first framing understates or disregards how technology can actually transform processes to create new ways of doing things, and the second glosses over the power of inertia among professionals that prevents them from proactively changing how they do things.
Robert – I read Jordan’s piece as saying three things: 1)AI can do a lot of legal tasks well; 2) Efficiencies and re-factored processes create opportunities to refocus where we prioritize our efforts and how we run our business; 3) Lawyers and law firm leaders that take those opportunities will redefine their operating models and definitions of success. Taken together, that’s transformational at a micro level, and as more lawyers and firms do likewise, transformational at the macro level of the profession. And that’s not even taking into account the impacts on the profession of institutions and clients to use their access to AI to dramatically change what they know, what they can accomplish without lawyers and what they expect of the lawyers with whom they engage. Not to mention the impact on business models and regulatory frameworks that follow from the enhanced abilities of people and organizations to serve themselves or to get legal support from unregulated or differently regulated service suppliers.
I read your piece as implying that those areas of legal practice not currently or expressly affected by AI will remain unaffected—whether directly or indirectly. “It looked that way a hundred years ago. It will look mostly the same in a hundred more.” I believe you are wrong.
The statistics, headlines and breathless press releases are only part of the story. I’ve spoken with many solo, small and large firm lawyers whose stories you won’t see in legal or general press that are already using AI in client and opposing counsel communication, crafting documents, pondering strategy and implications, supplementing or testing their intuition, coming up with ways of framing claims and defences, and improving the quality of advice they provide. For some, the way they practiced at the end of 2024 was already very different than the way they practiced at the beginning of 2024, and their appetites to go further are growing because, to a person, they are creating and spending their “AI dividends.” Many of them take pleasure in the fact that most of their contemporaries either haven’t noticed or simply believe it’s not possible.
Returning to where I started, if we only consider AI’s potential impact on the practice of law as coming from “easy button” products that fit into existing modes of practice and workflows, we’re missing out on the actual change already happening and will be caught flat-footed when the rate of change accelerates.
Have spent little time using AI. It seems fast and convenient. But I find myself always wondering about its answers. Not sure I’d ever be comfortable accepting its research findings on major issues or questions without doing my own. Agree with author’s assessment.